Introduction

Although fertility intention is a very good—in fact, arguably the best—predictor of childbirth at the individual level (Schoen et al., 1999; Toulemon & Testa, 2005), still we are well aware of serious mismatches between intention and realization. This is particularly true in the case of short-term, time-dependent intentions (Dommermuth et al., 2015; Régnier-Loilier & Vignoli, 2011; Spéder & Kapitány, 2009), as well as in the case of lifetime (family size) intentions and individual outcomes (Morgan & Rackin, 2010). The mismatch between intended family size (or any related ideals of family size) and fertility is also clearly visible when aggregate, country-level measures are compared (Goldstein et al., 2003; Ní Bhrolcháin et al., 2010; Sobotka & Lutz, 2011). Correspondence is somewhat stronger when the ideal or desired number of children across different cohorts is compared with completed fertility (Livi Bacchi 2001; Beaujouan & Berghammer, 2019); and we also know that this closer relationship of the cohort measures is, to some extent, a result of the over- or under-achievement of individual intentions (Morgan & Rackin, 2010). Yet, the study of intention remains at the forefront in terms of understanding fertility behaviour (Berrington, 2021; Lutz, 2020; Philipov, 2009a), and we believe that studying the correspondence between short-term intentions and the realization of those intentions can contribute especially to an understanding of reproductive decision making (Liefbroer et al., 2015).

Analysis of the link between intention and realization at the individual level reveals that the discrepancy is not unusual, and several factors influencing the link have been identified. Besides biological and emotional factors (Ajzen, 1988), the dynamics of partnership relations or unexpected life-course events may modify intentions and lead to their postponement or abandonment (Liefbroer, 2009), with the consequence that people forgo having children. Furthermore, several socio-demographic factors—such as partnership status, parity, age, labour market conditions or income—as well as cultural specificities and perceived normative conditions, may also facilitate or hinder the realization of intentions (Kuhnt & Trappe, 2016).

Studies that analyse several countries in parallel conclude that the overwhelming majority of those micro-level factors that influence the risks of realization operate in a similar way within the countries concerned; but they also reveal some country-specific factors at work (Régnier-Loilier & Vignoli, 2011; Spéder & Kapitány, 2015). Moreover, country comparisons highlight the fact that there is considerable country-specific heterogeneity in the rate of realization. However, these analyses do not address the characteristics of individual countries and societies that may cause the differences. The study reported here contributes to the literature of fertility intention in two ways. First, by using a rigorous procedure of variable construction we can reveal the heterogeneity in the rate of realization across 11 European countries, based on individual follow-up data. Here, we devote particular attention to the fact that the rate of realization is highly time dependent (Dommermuth et al., 2015; Schoen et al., 1999), and an accurate comparison can only be made if the time elapsed since the measurement of intention is exactly the same in all cases. Second, we aim to identify macro-level conditions that enable or inhibit the realization of short-term fertility intentions. Conceptual considerations will highlight the domains of social and economic dynamism, welfare state involvement and the ideational condition of the different countries. Individual factors are certainly included in our modelling, but only as control variables that enable us to uncover macro-level influences; therefore, micro-level results will not be discussed in detail.

To achieve our task, we utilize the first two waves of 11 countries in the Generations and Gender Survey (GGS) (Vikat et al., 2007). The dataset is unique in the analysis of population processes; and it is especially appropriate for the investigation of short-term intentions and their realization, since it includes a question that looks ahead (intentions over the next three years), and the succeeding waves of the survey enable changes in the various life domains to be measured. For our purposes, if there is a stated intention of having a child within three years, we are able to measure whether a child is actually born within that three-year window.

We proceed as follows. First, we provide an overview of the relevant literature and identify the existing research gap. The following conceptual framework includes a short overview of the theory of planned behaviour, highlights potential macro-level conditions and mechanisms, displays macro-level measures and formulates hypotheses accordingly. Then, some basic characteristics of the countries under investigation are outlined. The section on data and methods details the analytical strategy, the variables and the rigorous way of defining the outcome variable. The results are shown in three steps: descriptive results, an overview of the effects of the individual control variables and a detailed report on the macro-level effects. In the final section, we discuss our findings, set out the limitations of our analysis and suggest further research.

Previous Research into the Realization of Short-Term Fertility Intentions

The factors that influence the discrepancy between fertility intentions and achieved fertility are a recurring topic of research in population studies. The recent growth in interest can be linked to the persistent low fertility of several Western European countries and the plunge in fertility in Eastern Europe; furthermore, the contrast between ideal family size and fertility achievement enables us to identify a window of opportunity for policy making (Goldstein et al., 2003; Philipov, 2009a). The literature is abundant, and it uses different kinds of measures of intention and outcome. We limit our overview basically to studies of short-term intentions and outcomes.Footnote 1

MICRO-Level Determinants

A quite extensive corpus of literature focuses on the individual determinants of the realization or non-realization of fertility intentions in a single country (Berrington, 2004; Dommermuth et al., 2015; Heaton et al., 1999; Kuhnt & Trappe, 2016; Morgan & Rackin, 2010; Pailhé & Régnier-Loilier, 2017; Philipov, 2009b; Schoen et al., 1999; Toulemon & Testa, 2005). Based on these studies, several common factors can be highlighted. Some comparative analyses also investigate individual determinants and highlight general and country-specific patterns of micro-level effects (Régnier-Loilier & Vignoli, 2011; Spéder & Kapitány, 2015).

Demographic factors (such as age, partnership status and parity) clearly influence the success or failure of realization. A cohabiting partnership proved to be a prerequisite for successful realization of the intention. In some countries, the form of partnership also matters: whereas in the US and Hungary, the likelihood of a married couple realizing their intentions is greater than if they merely cohabit (Heaton et al., 1999; Schoen et al., 1999; Spéder & Kapitány, 2009), in France and Norway there is no tangible difference (Dommermuth et al., 2015; Testa & Toulemon, 2006). People who are in a ‘living apart together’ (LAT) relationship or who live alone have the lowest chances of realizing their intentions. The findings that relate to age highlight the ‘ticking of the biological clock’: women in the later phase of their fertility life course—usually those over 34—are less likely to realize their childbearing plans (cf. references above). Additionally, in some countries (the Netherlands) younger women often put off having children. Lastly, parity also has a powerful influence on realization: people with one child are more likely to realize their intentions, whereas people with zero parity are, in many countries, typically non-realizing postponers.

The influences of labour market position and the division of household chores are less clear cut, and gender-related country-specific conditions may play a bigger role (Hanappi et al., 2017; Kuhnt & Trappe, 2016). Women’s part-time employment often supports realization (Kuhnt & Trappe, 2016), whereas their unemployment hinders it (Pailhé & Régnier-Loilier, 2017). The link between full-time employment and realization may be influenced by whether the work is in the public or the private sector, or by other features of the job (Régnier-Loilier & Vignoli, 2011). The effects of the labour market position of the male partner are clearer: well-integrated, full-time participation in the labour market supports the realization of fertility intentions (Kuhnt & Trappe, 2016). A detailed and sophisticated analysis of the impact of the household division of labour, using pooled data from several countries, concludes that it affects the formation of intentions, but not their realization (Riederer et al., 2019).

The role of the level of education appears to be mixed. In France, better-educated women have a greater chance of realizing their intentions (Testa & Toulemon, 2006), whereas in Norway the effects are neutral (Dommermuth et al., 2015). By contrast, in the US the relationship is negative (Heaton et al., 1999; Morgan & Rackin, 2010; Schoen et al., 1999). Income effects are rather similar: women in higher income brackets are more likely to realize their fertility intentions (Berrington, 2004; Dommermuth et al., 2015; Hanappi et al., 2017; Schoen et al., 1999).

Finally, perceived norms, attitudes and perceptions also matter. That said, some variety in the results is apparent—partly due to the inclusion of different measures in the research programmes. Gender role attitudes, for example, influence realization in the US and the UK: in the US, women who profess traditional family attitudes become parents as they intend (Heaton et al., 1999); meanwhile in the UK, more career-oriented women aged over 34 have a clearly lower likelihood of realization (Berrington, 2004). Subjective norms have a significant effect in Germany (Kuhnt & Trappe, 2016) and also in an international comparison (Spéder & Kapitány, 2014). Those who state that their ‘significant others’ expect them to have a child stand a greater chance of realizing their intentions than those who do not. An increased feeling of uncertainty among highly educated Swiss women hinders realization (Hanappi et al., 2017), while an optimistic view of the life course among Hungarian men supports the realization of their intentions (Spéder & Kapitány, 2009). Satisfaction with the couple’s relationship proved to support a higher realization of intentions (Riederer et al., 2019). Lastly, the intention of having a larger family contributes to Norwegian people realizing their short-term intentions (Dommermuth et al., 2015).

Comparative Studies

Comparative research, based on individual panel data, has found significant country-level variations in the realization of short-term fertility intentions, especially between Western European and Eastern European countries (Bradurashvili et al., 2011; Riederer & Buber-Ennser, 2019; Spéder & Kapitány, 2014). Of those who planned to have a child within three years, two fifths actually succeeded in France and Germany, a third in Hungary and Georgia, and a fifth in Bulgaria. Furthermore, when a two-year time window between intention and realization is used to compare four countries (the Netherlands, Switzerland, Hungary and Bulgaria), a clear East–West divide is visible (Spéder & Kapitány, 2015). Lastly, when multivariate modelling is used, the difference between Western European and Eastern European countries is vast: the chances of individuals realizing their childbearing intentions in post-communist countries are less than half those of people in Western countries.

Taking account of differences between Western and Eastern European countries, and dividing countries into two corresponding groups, provides an opportunity to reveal contextual factors that shape urban–rural differences in realization, with higher realization rates in rural areas (Riederer & Buber-Ennser, 2019). Specific contextual factors related to urban–rural differences are clearly related to realization; higher female employment and higher childcare provision both support higher realization, but also contribute to greater abandonment of intentions. Overall, the study found that the effects of urban–rural differences decrease sharply, but do not disappear once contextual factors are controlled for.

Other country comparisons blur the East–West differences somewhat. On the one hand, such differences are sometimes only minor or quite negligible—e.g. when the capitals of Austria and Hungary are compared (Riederer & Buber-Ennser, 2018). On the other hand, certain differences can also be identified between individual Western European countries (Switzerland has a lower rate of realization than the Netherlands) and between various Eastern European countries (Bulgaria and Russia have lower rates of realization than Hungary) (Spéder & Kapitány, 2015).

Comparative analyses based on cross-sectional data to estimate the correspondence between short-term childbearing plans and births at the country level have also shown considerable country-level variation (Harknett & Hartnett, 2014). The level of realization (‘achievement rate’) seems to be generally higher when a pseudo-panel design is employed, than when an individual panel design is used. (The rate of realization based on cross-sectional data is estimated at around 50% in the 22 European Social Survey (ESS) countries.)

While there is a wealth of research on the individual factors that influence the realization of fertility intentions, and numerous studies that identify cross-country variations and commonalities in individual factors, no one has focused exclusively on the potential macro-level factors that shape cross-country variation. This is precisely the research gap that our analysis seeks to fill by exploring possible macro-level conditions and mechanisms that support or hinder the realization of fertility intentions. In this work, we attempt to distance ourselves from the separation of Europe into Western and post-communist (Eastern) countries, as this type of division can hamper the exploration of more general patterns (Müller, 2019). While past traditions certainly live on, by time the data analysed were collected, most of the former socialist countries were already members of the European Union and part of the European, global economic order.

Conceptual Framework: Macro-Level Conditions with the Potential to Support or Hamper Realization

Before reviewing potential conceptual approaches and research findings that highlight the macro-level conditions and mechanisms that support or hinder the realization of fertility intentions, it may be helpful to provide a brief overview of the theoretical framework used for intention formation, as this will help locate the potential macro-level effects discussed below.

A Sketch of the Theoretical Framework for Intention Formation and Realization

The Theory of Planned Behaviour (TPB) developed by Ajzen (1988) and used by the GGS is a social-psychological action theory that places great emphasis on understanding which factors and mechanisms influence the formation of intentions, and that considers the relationship between intention and behaviour to be relatively straightforward. Ajzen describes it as follows: ‘Intention is … assumed to be the immediate antecedent of behavior’ (Ajzen, 2002: 179). There are three subjective (yet different) types of factors that determine the formation of intention. Attitudes relate to the given object and to evaluations about the expected outcomes (advantages and disadvantages) of the behaviour. Subjective norms include expectations that arise from the network of ‘significant others’. Perceived behavioural control covers those factors that may make realization of the given behaviour easier or harder. The three influencing factors are weighted according to the importance ascribed by the individual concerned, and all are based on beliefs that are shaped by background factors not detailed further here (Fig. 1; for an overview, see Ajzen, 1988; Ajzen & Klobas, 2013).

Fig. 1
figure 1

A schematic representation of the Theory of Planned Behaviour

Full size image

The actual short-term fertility intention is a result of careful deliberation: people at a given point in time consider the particular aspects of behaviour (including the financial and emotional advantages and disadvantages); how important particular aspects are to them; the importance that they attach to the expectations of significant others regarding childbearing; and finally, how they view perceived barriers to childbearing.

In the original formulation of the theory, only limited attention is given to factors that ‘can disrupt the intention–behavior relation’ (Ajzen, 1988: 132). Emotions, dependence on others and unforeseen life-course events may all frustrate the realization of the intended behaviour. In a paper, the designer and protagonists of the TPB theory applied it to fertility behaviour (Ajzen & Klobas, 2013), highlighting in particular the role played by enablers of and constraints on behaviour—for example, the lack of available childrearing institutions. The framework is also very helpful in considering potential macro-level factors and trying to locate them. Here, we can especially consider whether a macro circumstance under discussion could modify any of the factors behind intention formation and change the stated short-term intention, or whether it could loosen or strengthen the intention–behaviour link.

Economic Dynamics and Uncertainty

Social change is inherent in modern society: change and renewal are constant features of Western society, and economic and social innovations play a key role in renewal of the social system. People act and take life-changing decisions (e.g. about having a child) in ‘customary times’ or ‘times of normality’ by taking account of their circumstances and any anticipated change to them. But the pace of social change may be altered (Zapf, 1996). For example, the Great Recession was unanticipated and brought with it a fall in fertility behaviour (Comolli, 2017; Goldstein et al., 2013). Similarly, the profound socio-political transition in Central and Eastern Europe, involving a shift from a redistributive economic system to a market economy and the associated transformation of the labour market, was also unforeseen. At times of unanticipated and profound change, life-altering decisions are avoided or postponed (cf. Rodin, 2011). Such societal conditions were stressed by Spéder & Kapitány, 2014 when they showed that the chances of individuals realizing their childbearing intentions in the post-communist countries were less than half those of people in Western countries. Considering potential factors influencing the different rates of realization, the unprecedented dynamism of the altered structural conditions during the regime change was highlighted by those authors.Footnote 2

Fertility intentions tend to be formulated in the context of social dynamics as currently anticipated, with the actors making an assessment (attitudes) and taking into consideration the circumstances that could help them or hinder them (perceived control). It is assumed that economic or societal upheaval will play a key role in the non-realization of people’s short-term fertility intentions, since unusual and unanticipated fluctuations may result in a revision of intentions and could create fresh impediments (actual control) that loosen the link between intention and realization (Ajzen & Klobas, 2013). In order to capture the potential influences of social and economic dynamics on childbearing decisions, some measure that reflects fluctuations on the labour market and the consumer market (prices) could be taken into account.

Employment uncertainty is a major obstacle to having a child (or another child). The unemployment rate—the key indicator of labour market fluctuation—is a proven macro-level factor that influences fertility (Comolli, 2017; Goldstein et al., 2013). Its dynamic nature is well suited to our analytical purposes. Since unemployment among people aged 15–24 may be more relevant to a generation that is starting to think about parenthood, and may be a more sensitive measure of labour market fluctuation, it is also worth considering youth unemployment. If the focus is on measuring unanticipated change, we can go even further and gauge the intensity of labour market change: an indicator of the youth unemployment swing reveals the maximum deviation of the unemployment rate from the average unemployment rate over a given period of time.

Rising consumer prices (inflation) affect the economic conditions of childbearing (Cornia and Paniccia 1996: 113–114). In simple terms, inflation increases the cost of having children, since the cost of goods and services for children rises disproportionately, eats into the savings of a household and often undermines the value of family benefits. Since ‘inflation per se is perceived as a serious source of instability by most economic agents’ (ibid: 113), we may assume that inflation is a relevant measure of uncertainty for the general public. The results of research that points to the effect of inflation on happiness (Di Tella et al., 2001) may also indirectly support our idea of treating inflation as an indicator of economic dynamism that has a bearing on the realization of short-term fertility intentions.

Overall, we assume that the greater the stability and the less fluctuation there is on the markets (i.e. a low unemployment rate, low inflation rate, lower swings on the labour market), the better are the chances of people realizing their short-term fertility intentions.

The Institutional Context: Comprehensive Social Protection or Spending on Families

From the point of view of fertility behaviour, family policy packages are of key importance; and sure enough, country studies underscore their effect on fertility (e.g. Hoem et al., 2001; Milligan, 2005). The results of comparative papers are more mixed, however. One comprehensive European comparative analysis distinguished between five types of family support measures and showed that all the measures had some effect on fertility in developed countries. However, it concluded that it was the specific ‘mix’ of measures that was most important (Luci-Greulich & Thévenon, 2013).

The effects of the welfare state may be more indirect. While it has various functions, we can highlight the fact that the influence and the effects of globalized markets are to some extent filtered by welfare state institutions, labour market regulations, etc. (Mayer, 2001; Mills & Blossfeld, 2005). Furthermore, welfare state packages provide a basic safety net for ordinary people, safeguarding them against labour market, health and income risks, and thus offering them a degree of stability (Leisering, 2003).

The formation of intentions takes place within the context of a particular (welfare) state and within a particular family policy setting: actors are aware of the specific family support that they can expect, of the institutional help available in bringing up children, etc.—although there can be a significant divergence in the information available to them, especially between those currently without children (Parity 0) and those with (higher parities).

How, then, does the institutional setting influence the realization of fertility intentions? Two types of influence are assumed. Obviously, an unforeseen alteration in the institutional system (such as a change in access to subsidies) may affect those attitudes that shape intentions or indeed modify the actual enablers of the behaviour, thereby loosening the intention–outcome link. On the other hand, the extent of welfare spending and family support may affect the likelihood of intentions being realized. In this regard, the coverage of the safety net and the family policy packages indicates the availability of state provision in the case of both generalized risk and life-course-related risk (Leisering, 2003; Mayer, 2001; Mills et al., 2005). Therefore, it can be assumed that the further the welfare state extends and the closer the welfare net is woven, the lower the risk from unanticipated social and economic change and the less the perceived uncertainty. Similarly, the extent of the family support system (i.e. government spending on families) shows the government’s commitment to contributing to the cost of children and may signal how far families can rely on state support in raising children.

It would be useful to take family policy and other institutional changes into account, but the lack of comparable policy indicators does not permit this. However, given the slow pace of institutional change, we can assume that the differences between countries will not be dramatically affected in the short (3-year) time period we focus on. (That said, the effects of policy change on the realization of intentions should be a subject for future research.)

Overall, we assume that the more generous the welfare state is, the greater the general social provisions are as a proportion of GDP; and the bigger the state’s financial involvement is in covering the cost of raising children, the greater the probability that childbearing intentions will be fulfilled.

Ideational Conditions: Insistence on Traditional Family Attitudes or Autonomy?

Highlighting values in shaping demographic behaviour, including childbearing behaviour, is nothing new. According to the Second Demographic Transition theory, the main driver of change in couple and childbearing behaviour is value change (van De Kaa, 1987; Lesthaeghe, 2010). Reher emphasizes the importance of cultural path dependence in understanding demographic behaviour (Reher, 1998). It is therefore reasonable to pay attention to whether, and how, value orientations and predominant attitudes can influence the formation of the intention to have children and the chances of achieving those plans.

According to our framework of intention formation (the TPB), beliefs about having a child play a crucial role in shaping attitudes, perceived norms and perceived control—and consequently, in the formation of intention. Individual beliefs are rooted in and shaped by the normative system of a country (Liefbroer & Billari, 2009; Mönkediek & Bras, 2017). The normative system includes, among many other things, norms concerning partnership and family forms, attitudes towards gender roles, desired family size, expectations about the timing and sequence of family events, and ‘mental timetables’ (Hagestad & Neugarten, 1985). Since people living in different countries of Europe clearly differ in terms of their family-related attitudes and gender role norms (Hagenaars et al., 2004; Lück & Hofäcker, 2003), it is justifiable to assume that these differences contribute to variation in intention formation. But the key question is how the dominant beliefs or orientation can influence the realization of short-term fertility intentions. Since value change occurs relatively slowly, and often via generational replacement, it would be misleading to assume that value change brings about change in short-term intention and contributes to lower realization. Rather, it is advisable to look for any mechanism that weakens or strengthens adherence to formed/stated intentions.

In reviewing the relevant literature, we found three mechanisms that could be relevant to our investigation. First, the strength of the intention may differ at the country level. It is known that the strength of attitudes at the individual level clearly differs depending on information about the related action, on the involvement of the actors, etc. (for an overview, see Krosnick & Petty, 1995). Since the strength of the attitude differs, the strength of the intention may also differ. Accordingly, we assume, for example, that in a country where the idea of the traditional family is dominant, adherence to fertility intentions is stronger. This could mean, when all else is controlled for—i.e. when differences in social dynamics are also considered—that the prevailing normative environment encourages those concerned to implement their intentions. Consequently, a greater proportion of them do so.

However, the dominance of family-related beliefs may support a different, perhaps opposite, type of reasoning: a society in which traditional values predominate may provide a stronger motivation to express the intention of having a child in the short term, since having a child is greatly valued in society. Then, in the face of changing living conditions, that intention may turn out to have been overoptimistic, in which case it is more likely to be revised. In this case, it is precisely those countries that favour more traditional family roles that experience greater revision (and non-fulfilment) of intentions. Weinstein’s concept of ‘unrealistic optimism’ (Weinstein, 1980) supports the above reasoning. According to this concept, people are often optimistic and sometimes pessimistic in their assessment of major life events (Shepperd et al., 2013). For example, college graduates may be optimistic that they will not become pregnant within the next year (Rothman et al., 1996). Could we not view the low level of attainment as a result of people’s optimism being supported by dominant views (perceived pressure) about their ability to have a child within the three-year period? The fact that unrealistic optimism may vary across cultures (Heine & Lehman, 1995) allows for cross-country differences in the intention–outcome relationship.

Next, the idea of response conformity (Bond & Smith, 1996) may indirectly support the last reasoning: a society in which traditional values predominate—or alternatively, a society where government and the broader public stress the long-term negative consequences of low fertility—may provide greater motivation (or pressure) for people to declare their intentions of having a child in the short term, since childbearing is highly valued in society. But when faced with reality, people may find that they have overstated their intentions—in which case they are likely to revise them. In this case, it is precisely in countries where people favour more traditional family roles that we see greater revision (and non-fulfilment) of intentions.

We can hardly claim to have accounted fully for the impact of potential mechanisms of dominant attitudes and beliefs on the realization of intentions. That is, there may well be other inhibiting forces and motivating mechanisms. In fact, we can only indirectly test the three mechanisms mentioned. While the first mechanism suggests that we can expect higher attainment where traditional beliefs are dominant, the last two suggest the opposite.

Identifying the impact of dominant ideas on realization is perhaps even more cumbersome than the assumptions discussed in the previous two sections, since there are no commonly agreed indicators available and we have to make do with the available measures of opinion from other research. It is also useful to take into account the principle of compatibility (Ajzen & Klobas, 2013: 208ff). Accordingly, it is not general value orientations, but attitudes—opinions and evaluations that are closely related to the object of investigation (in our case, having a child)—that are potential factors influencing the realization of intentions. Taking account of potential and available measures of related dominant ideational conditions, three indicators are examined in terms of the extent to which they influence adherence to stated intentions and the fulfilment of those intentions: (a) the acceptance of non-traditional family forms; (b) the perceived importance of a child for a woman’s sense of fulfilment; and (c) the question of whether having a child is or is not a purely private matter.

Some Features of the 11 Countries Analysed

A detailed report on the countries analysed is beyond the scope of our study, but a few basic indices (GDP, total fertility rate (TFR), religiosity) suffice to illustrate the diversity of the 11 countries (see Table 1)—although the picture would certainly be even more varied if Southern European countries could be added. Differences between the Western and the post-communist Eastern countries are frequently cited.Footnote 3 Several key macro-level indicators suggest considerable heterogeneity across countries—for instance, in terms of per capita GDP or the prevalence of extra-marital births. Meanwhile, however, the TFR generally shows no marked differences—and particularly not among the post-communist countries.

Table 1 Selected economic, social and demographic indicators of countries, 2005
Full size table

In terms of economic performance (GDP per capita), the four Western European countries have similar welfare levels, which far outstrip those of the post-communist countries. But among the latter, there are some marked differences: for instance, Georgia’s economic performance is only about a quarter of Hungary’s. Measured in purchasing power parities (PPP) (over the period 2000‒2005), the developed Western countries demonstrate economic performance roughly eight times that of Georgia.

In terms of the TFR—the key measure of fertility—postponement and changes in the family pattern mean that all the countries surveyed in 2005 (except France and Sweden) had a very low figure, of around 1.3. After that, up until the Great Recession the TFR rose somewhat in most countries. Of the Western countries, France and Sweden had very similar levels of fertility, at around the replacement level. The German-speaking countries have displayed low levels of fertility for several decades—for reasons other than changes in the fertility model. Some point to a bifurcation scenario (Rindfuss et al., 2016), while others (such as Sobotka, 2016) emphasize a similarity between Austria and Czechia. (High levels of childlessness in Germany and Austria contribute greatly to their low fertility rates.)

The goal here is not to provide an inclusive or comprehensive picture of each country’s conditions. However, it is useful to show a range of differences in European economic performance, institutional systems and cultural climates. It also points to the demographic and social conditions to which our analysis refers.

Data and Methodology

Data, Sample and Attrition

Our analysis is based on data from the Generations and Gender Survey (GGS), which captures the dynamic features of demographic behaviour by collecting longitudinal data (Vikat et al., 2007). The GGS is a follow-up study: sample members are interviewed at three-yearly intervals. Our analysis takes into account every European country for which data are available for the first two waves of the GGS. In the countries under consideration, the first interview took place generally in the first decade of the century,Footnote 4 with the second following (roughly) three years later.

In order to ensure an identical age range for respondents in all countries, only those aged 21–45 at the time of the first wave were included in the analysis.Footnote 5 Pregnant women and men with a pregnant partner at the time of the first wave (defined on the basis of the woman having given birth within 6 months of the first interview) were excluded. In accordance with our research question, all women and men who responded positively to the intention question (‘Do you intend to have a/nother child within three years?’) were included in the analysis. (The two positive answers (definitely yes, probably yes) were collapsed into one (yes), since for Hungary we had only the ‘yes’ answer.) Altogether, 8886 respondents from the 11 countries intended to have a child within three years. Since it is essential to include the woman’s characteristics (age, labour market status) as control variables in the multivariate analysis, every woman is included in the analysis, but only those men with a co-resident female partner are included. Consequently, our working sample has N = 6498.Footnote 6

Attrition is a standard feature of follow-up studies, but the concern rises when attrition is biased. The risk of bias increases as the attrition rate increases. In our case, three of the 11 countries—Czechia, Lithuania and Germany—have unusually high attrition rates; seven have attrition rates that are not unusual for panel studies; and in one case, Sweden, there is no attrition.Footnote 7 Few analyses discuss dropout in the GGS (Bartus et al.,2015; Brzozowska et al., 2021), and they tend to focus on a single country (Buber-Ennser, 2014; Régnier-Loilier and Guisse 2012). Overall, they conclude that analyses using second-wave data do not lead to biased results in fertility analyses, although only one analysis includes a country (Germany) with a very high dropout rate.

To strengthen the basis of our analysis, we conducted an investigation of panel continuation, using logistic regression. The selection of influencing factors was based on the relevant literature (Watson & Wooden, 2009). For the purpose of our analysis, we examined two questions: (i) whether the dropout factors differ between countries with low and high dropout rates, and (ii) whether there is a relationship between responses to the childbearing intention question and dropout rates, as this may affect the relevance of the responses to our research questions. The analysis was conducted on a sample designed according to the criteria of our research question, which included women aged 21–45 who responded in the first wave and men aged 21–45 with a female partner living with them. As a result of the analysis, we find that women, older respondents aged 34–45 (and partly 29–33), women with a cohabiting partner, those with two or more children and those with higher education are more likely to be willing to respond in the second wave (see Table 6). In some countries, scarce financial resources are a barrier to responding in wave 2. Importantly for our analysis, in nine countries with attrition there is no relationship between the variable of intention to have children and willingness to respond. Only in Germany do we find that those who planned to have children in the short term were more likely to respond to wave 2 questions. In the three countries with a very high dropout rate and in the seven with an average dropout rate, the factors influencing panel participation are consistent. Where gender, age, partnership, number of children and specific categories of level of education have a significant effect on retention, their effects are in the same direction in the two country groups.Footnote 8 All this leads us to conclude that, although the fact that three countries had a very high dropout rate remains a cause for concern, the coincidence of factors shaping dropout mitigates this concern. It also provides a good basis for including countries with a high dropout rate in our multivariate analysis.

Measures

Dependent Variable

A valid comparison of fulfilment rates requires identical measures of intention (having a child within three years), fulfilment (childbirth) and time elapsed between expression of intent and realization. Since the timing of the second wave and the length of the second-wave fieldwork varied somewhat across the countries, and since the degree of fulfilment of fertility intentions clearly depends on the time elapsed since the intention was measured (Davidson & Jaccard, 1979; Dommermuth et al., 2015; Schoen et al., 1999), a rigid time window—the ‘time at risk’—had to be defined. A birth is considered to have been the realization of an intention if it occurred in the period of the 7th to the 36th month following the first survey, when the intention (‘want to have a child within three years’) was measured. If no birth occurred during the 7–36-month period, non-realization was stated. Note, the ‘time at risk’ is thus the same in each country.

Individual Control Variables

The selection of our individual control variables is based on the findings of earlier studies and takes account of the potentials and limitations of the comparative datasets. The variables of sex, the woman’s age group, partnership status and the woman’s labour market status (and the various categories) are self-explanatory (Table 7 in the Appendix). But two variables require some elucidation: since a readily comparable indicator of income status is lacking, income position is replaced by perceived income position (i.e. the perceived income needs of the household budget (‘making ends meet’)). The values of this variable indicate whether the household is making ends meet (i) easily, (ii) with some difficulty, or (iii) with great difficulty. The perceived subjective norm index measures the extent to which the respondent feels that ‘significant others’—parents, friends, relatives—expect the respondent to have a child (a lower value indicates a higher expectation). Reference categories are given in Table 7 of the Appendix, where the variables are listed.

Country-Level VariablesFootnote 9

According to our conceptual framework, four measures of economic and social dynamics are included.Footnote 10 The unemployment rate and the rather more volatile youth unemployment rate (15–24), measured at the time of the first interview, are well-known measures of economic fluctuation (i.e. recession and prosperity). (See Table 2 for values of macro-level variables.) The ‘swing’ in the unemployment rate seeks to capture the magnitude of the change in the youth unemployment rate. The relevant indicator is derived from two measures that cover a nine-year periodFootnote 11: i.e. the maximum differenceFootnote 12 in the youth unemployment rate is related to the average youth unemployment rate over that period.Inflation aims to capture economic and social uncertainty and unpredictability: the inflation rate is measured at the time of the first wave (actual inflation).Footnote 13

Table 2 Country values of macro-level variables
Full size table

Welfare state involvement is measured by two closely related indicators: total social expenditure, as a percentage of GDP; and social protection spending on children, again as a percentage of GDP (ILO, 2017: 402–413).

Three measures seek to capture the prevailing ideational climate related to family and childbearing. One—often found in related analyses (Thornton & Philipov, 2009; Thornton & Young-DeMarco, 2001)—is a person’s attitude toward the institution of marriage (level of agreement and strong agreement with the statement ‘Marriage is an outdated institution’). The second is whether a person believes having children is a private matter (i.e. is exclusively related to the sphere of individual autonomy) or whether there are communal obligations (norms) involved as well (level of agreement with the statement ‘People should decide for themselves to have children’). And the third measure likely to indicate how central the role of children is to the life of a woman gauges the level of agreement and strong agreement with the statement ‘A woman has to have children to be fulfilled’. In each of these attitudinal measures, the percentage of people who agreed with the relevant statement became the country-level measure.Footnote 14

Analytical Procedure

A multilevel binary logistic regression model was employed to model the realization of fertility intentions in 11 European countries, using the pooled dataset. Country-specific individual-level data typically have a multilevel structure, since subjects within the same country may have outcomes that are correlated with one another, due to the similarity of the general context. The conventional single-level logistic regression is unable to account for this kind of intra-cluster correlation. Furthermore, ignoring the multilevel structure of data can result in biases in parameter estimates and their standard errors. By taking account of the correlation within the cluster, we are able to make reliable parameter estimates of within-country effects.

We used random intercept logistic regression models. The model derives its name from the fact that the intercept is allowed to vary randomly across countries, through the introduction of cluster- (country-) specific random effects. The estimates of the extent of the similarity of subjects within a country can provide an important insight into the group-level effects on individual fertility behaviour.Footnote 15 Moreover, in accordance with our primary interest here, we extended our models by adding country-specific attributes to measure explicitly the size of the effect of different structural conditions.

We are aware of the problem of having an extremely small number of cases at the country level (Level 2), which can lead to estimation bias, as discussed in the literature (Bryan & Jenkins, 2016). However, we accept and prefer the argumentation of Robson and Pevalin (2016), who contend that ignoring the group variance may lead to a bigger error (ibid: 27). Note that in the case of some important comparative datasets (e.g. Survey of Health, Ageing and Retirement in Europe (SHARE)) there are also around a dozen countries, and multilevel models are used successfully (Engelhardt, 2012). Additionally, alternative analysis (results not shown) strengthened the relevance of the analytical approach used. A country analysis of the realization of intention confirms the inclusion of the individual (Level 1) factors selected for the multilevel modelling; an analysis with country dummies using the pooled data supports the assumption of significant country differences.Footnote 16

We will do our modelling step by step, always including only one macro variable in the multilevel models (with the exception of the final model, which will include two). The goodness of the models will be assessed by evaluating the between-country variance and the inter-class correlation (ICC). Both show whether a multilevel model is worth employing, and whether the introduction of the given country variable reduces the error term. The between-country variance shows the effect of country-specific predictors that have not been controlled for. Therefore, a significant decrease following the introduction of a country variable indicates that the given Level 2 factors have a sizeable effect on realization. The ICC represents the ratio of the unexplained variance in the country level to the total variance. If it is close to zero in the empty model, then there is no sense in using the multilevel model. On the other hand, if the introduction of a macro variable causes the ratio to decrease markedly, then the variable in question contributes significantly to the explanation. Lastly, the Akaike Information Criterion (AIC) measure reports on the fit of the model—something that is relevant if we are comparing two models. A clear reduction in the AIC signifies a more robust model.

Results

Descriptive Results

The rate of intention realization differs greatly across the countries when positive intentions are compared (Fig. 2). A glance at the length of the bars makes it immediately clear that the differences are large and significant: two fifths of short-term childbearing intentions in Germany, France and Sweden (39‒41%) were realized, but the figure was less than a fifth in Bulgaria and Russia (15.7% and 17.0%, respectively). It is interesting that three countries with quite different social systems—Austria, Czechia and Hungary—exhibit very similar levels of fulfilment. Here it should be noted that the two categories definitely yes and probably yes were collapsed. Generally speaking, if those two categories were treated separately, the rate of realization among those who responded definitely yes would be higher in each country, but the country heterogeneity and country rankings according to realization would remain the same (see Table 9 in the Appendix).

Fig. 2
figure 2

Ratio of those intending to have a child within three years who actually did so within 7–36 months (European countries, all females aged 21–45 and partnered males aged 21–45) Source: own calculations, GGS first and second waves (first-wave country weights are used)

Full size image

Since realization is dependent on individual characteristics (see below), country differences in terms of fulfilment are partly due to the different proportions of specific groups (compositional effect). Thus, for example, if in a particular country there are lots of women or men who live alone or in a LAT relationship, and who intend to have a child within the next three years, that will depress the country’s fulfilment rate, since the probability of realization within those groups is very low. The countries examined clearly have a higher fulfilment rate if co-resident people or those with one child (Parity 1) are considered. For example, if we focus on cohabitees from the first wave, then half of all intentions in France and Sweden (49.7% and 49.4%, respectively) were realized; the figure in Russia was still only a fifth (18.9%) and in Bulgaria a quarter (23.1%). Hungary is in an intermediate position, as 29.1% of cohabitees who intended to have a child actually did so. If a social group with high fertility risk is selected, realization lies between 35 and 66%, while the heterogeneity of countries persists. (For the proportions of all the countries and specific social groups examined, see Table 10 in the Appendix.)

Our findings—which reveal large country variation in the realization of short-term intentions—are consistent with results from other individual follow-up studies. Analyses comparing Western and Eastern European countries almost all show higher realization rates for Western European countries. The exception is a comparison between Austria and Hungary, where very similar rates are found. Notably, the results based on GGS are consistent with our earlier analysis using non-GGS data and a two-year intention time window (comparing two Western European and two Eastern European countries in terms of intention realization within two years). Based on the non-GGS data analysis, the Netherlands had an exceptionally high realization of intentions within two years (75%) and Bulgaria a low realization (38%). It is worth noting that in this case the rate of realization was clearly higher on account of the narrower time window covering intention and outcome.

However, our estimation results, which are based on individual data, do not support estimates based on cross-sectional data, which take group characteristics into account. Harknett and Hartnett (2014) used data from the second and fourth waves of the cross-sectional European Social Survey (ESS) data collection to estimate the realization of fertility intentions within three years: intention was measured on the basis of the 2004 data and actual births on the basis of the 2008 data. Using the data from these two waves, an ‘achievement rate’—the proportion of children born within three years, relative to the proportion intended—was estimated for each country. The authors were aware that these ‘achievement rates’ were higher than the realization rates, since their estimates also included unplanned and ‘sooner-than-intended’ births. Since information on these rates was available for only a few countries, they estimated the average rate of European realization at around 50%, some 10% lower than the ‘achievement rate’ (60%). The 50% average is clearly higher than the average could be on the basis of an individual follow-up survey such as the GGS.

However, Brzozowska et al. (2021) show that there is a very significant difference between countries in terms of the proportion of unplanned births and of children born earlier than intended. Of all the unplanned or sooner-than-intended births that occurred between the first two waves of the GGS, France took a share of 8%, Poland 30% and Hungary 34% (Brzozowska et al., 2021). In the light of the analysis of unplanned and sooner-than-intended births, an estimate of the ‘achievement rate’ that is based on cross-sectional data cannot be considered a good approximation of a country’s realization rate.Footnote 17 The variability is probably also the reason why France, which has a high realization rate according to the GGS (cf. Régnier-Loilier & Sebille, 2016), has—according to the ESS estimate—a lower ‘achievement rate’ than Hungary, which has a low realization rate based on the GGS and a high rate of unplanned or sooner-than-intended births.

Finally, of course, we cannot ignore the unreliability of some GGS country estimates due to the high dropout rate, since three countries have very high attrition rates. Thus, although the dropout data do not appear to be biased in terms of childbearing intentions, our estimation results for countries with a high dropout should be treated with some caution.

Impact of Individual-Level Characteristics on the Realization of Fertility Intentions

Since the parameters of the individual variables are very stable across all models, and barely change when we include different macro-level variables, we discuss the influence of individual factors using the parameters of the multilevel random intercept model without a macro covariate (Table 3).Footnote 18 If relevant, the results from models with macro variables (Table 11 in the Appendix) are mentioned.

Table 3 Odds ratios of realizing short-term fertility intentions, effects of individual characteristics (results of multilevel random intercept model without macro covariates, M1 model)
Full size table

Overall, the majority of the associations are in line with the earlier research results outlined above. For a woman in the latter half of her thirties, the chances of realization clearly decline: those aged over 35 have only half as much chance of fulfilling their intentions as 29‒34-year-olds. Furthermore, in several models, when macro variables are included, women aged 24–28 have a greater chance of realization than the reference age group. (Note, this statement is somewhat weaker when significance levels are considered.) Partnership clearly counts: cohabiting couples have two and a half times more chance of realization as people living alone or in a LAT relationship. As far as parity is concerned, women with Parity 2 or more have significantly less chance of having the intended child than do women with lower parities. A comparison of zero parity and Parity 1 shows that women with Parity 1 have a somewhat better chance of realization, which is in line with the literature reviewed. (In those models with macro variables (see Table 11), the coefficients of Parity 0 are always higher than 1, but the effects are only significant in some models—and then at a very low level.) Considering the labour market position of women, inactive women have a clearly lower chance of realization than unemployed (and employed) women (cf. Table 3 and Table 11). Comparing employed and unemployed women, the former have a somewhat lower chance of realization, although the association is valid at a lower level of significance. Of course, caution should be exercised here, as labour market regulations may be country specific. Subjective norms are significant in all models: those that have a sense of greater normative expectations are more likely to realize their intentions. Lastly, subjective income level has an effect in line with the expected direction: economic hardship (major difficulty) hinders the realization of intentions. As we showed in our earlier studies, subjective (self-assessed) income is mainly important in the former communist countries (Spéder & Kapitány, 2014). Overall, the individual factors show great stability across the modelling (they are identical to the third decimal place); thus, we will not bother to show them when we present the macro-level effects of the multilevel models.

Effects of Macro-Level Factors

The ensuing multilevel models reveal how the macro-social conditions characteristic of specific countries affect the fulfilment of fertility intentions. We start with the empty model (Model 0) without any covariate, but assuming the multilevel structure of the data. Model 1, which includes only individual variables and assumes also the multilevel structure of the data, is a key reference, since this model controls for country differences that are due to compositional differences regarding individual factors. Then, step by step, we introduce the macro variables (Level 2)—always just one at a time—to see whether and how they influence the fulfilment of fertility intentions (see analytical procedure section). Lastly, we present a model where two macro variables are included.

Generally, we are interested in whether the country-specific (Level 2) variables introduced have a significant effect on the realization of fertility intentions. In parallel, we consider two measures—between-country variance and related inter-class correlation (ICC)—and see if they indicate whether the given multilevel model reduces variance, and if the introduction of country variables is statistically significant.

The empty model (Model 0) and the model of the individual variables without any macro covariate (Model 1) serve as benchmark models (see Table 4). This means we will compare all models with macro variables against Model 0 and Model 1. The ICC of our empty model, which measures the share of variation attributable solely to country characteristics, is above 0.05—that is, according to the rule of thumb, the cut-off point for using multilevel models. However, if we include the individual variables (Model 1), the ICC increases to 0.079. This reveals some slight compositional effects. The fact that in Model 1 the between-country variance and the ICC are clearly larger than in the empty model tells us that, if individual effects are controlled for, there are greater country differences than in the empty model.

Table 4 Odds ratios of realizing short-term fertility intentions, parameters of macro-level characteristics (results of multilevel random intercept models, individual-level variables controlled for)
Full size table

Now, pursuing our main interest—namely, how the macro indices affect the fulfilment of intentions—we consider the effects of the different country factors. On the one hand, six of the nine variables show significant effects, but reduce the between-country variance to differing degrees (see Table 4). On the other hand, the macro indicators included do not always have the expected effects.

The unemployment rate—one of the most reliable macro indicators in explaining macro-level fertility change—does not influence the fulfilment of fertility intentions, irrespective of whether the general rate or the youth unemployment rate is used. However, a specific aspect of unemployment—the swing (amplitude) in the unemployment rate—does seem to affect the realization of intentions. Comparing Model 1 and Model 4, we see that the ICC is reduced by 40% (from 0.079 to 0.047). The parameter of the unemployment swing—an odds ratio lower than 1 (0.279)—tells us that the greater the swing in the youth unemployment rate of a country, the lower the chances of short-term fertility intentions being realized. While the unemployment rate does not have a direct effect—perhaps because it is part of the social and economic context when the intention is formed—big changes in youth unemployment may signal instability or volatility, and may lead to a revision of intentions. Overall, stability on the labour market fosters the realization of fertility intentions, whereas vast change and instability hamper it. The reduction in the ‘between-country’ variance is greaterFootnote 19 if inflation is included in our model (Model 5). The ICC—the rate of unexplained country-level variance—more than halves, decreasing to 0.036. According to what we see, the higher the inflation, the lower the likelihood that short-term fertility intentions will be realized.

Both the indices that measure welfare state involvement have a significant effect. Since they are correlated, it is not surprising that their effects should operate in the same direction and should be relevant in the same way. Both total social expenditure as a percentage of GDP and spending on children facilitate the realization of intentions. The higher the expenditure as a percentage of GDP at the country level, the greater the chances of realization. Total social spending as a proportion of GDP has a statistically stronger effect than spending on children, since it reduces between-country variance. Total social expenditure has overall the lowest ICC (0.025) of all the models with one macro variable compared here. We assumed that spending on children would have a stronger effect, since it has been proved to influence TFR (Luci-Greulich & Thévenon, 2013); in the case of the realization of intentions, perhaps the general risk-covering function of the welfare state is more important than the cost-reducing function of spending on children.

As for dominant ideational (attitudinal) characteristics, the results are mixed. The variable measuring support for marriage, often used as an indicator of traditional views in a society, displays no significant effect on the fulfilment of fertility intentions. However, the two indices attached to having a child do have an effect: there is a greater chance of someone having a child (in fulfilment of the intention) in those countries where the prevalent view is that the decision to have children is entirely a private (individual) one (i.e. where fewer people believe that having children is also a collective obligation). Furthermore, the proportion of intention fulfilment is greater in those countries where fewer people believe that having a child gives meaning to a woman’s life. The results do not support our initial assumption that in a society with predominantly traditional views, people will stick more closely to their initial intentions. On the contrary, realization is higher when the dominant view is that having a child is a private matter and not a normative obligation.

We should be cautious in assessing which model seems most relevant in explaining the realization of fertility intentions. Based on the AIC, the model including total social expenditure as a proportion of GDP (Model 6) is the best, followed by the model featuring inflation (Model 5) and the model that includes the prevailing ideas about whether having a child is a private matter (Model 9). But note also that there is significant correlation between the rate of social expenditure and the notion that having a child is a matter for the individual. We do not assume a causal relationship between the two factors, and nor do we assume a dependence on a third country-level factor; but the last cannot be excluded.

Four macro variables that were found to be significant were selected and their models were compared to identify the best fit with our data structure. We calculated several Goodness of Fit measurementsFootnote 20 and, based on these, we found that the model with social protection (Model 7) was the best, followed by the model with inflation (Model 5) and the model that considers the notion of whether having a child is a private or a public matter (Model 9). The model with the least-good fit was the ‘unemployment fluctuation’ model (Model 4).Footnote 21

The small number of countries (Level 2) and possible associations between the potential macro-level variables both prevent us from including several different combinations of the Level 2 variables in the models. Nonetheless, we experiment by including two unrelated country variables at the same time. Model 11 pairs inflation and the variable for support for the view that having a child is a private matter (see Table 4).

Statistically speaking, the model improved markedly (between-country variance is 0.051, ICC = 0.015) and both variables remained significant, while the signs of the parameters are as before. The lower the inflation rate (the less the uncertainty) in a given country and the stronger the view there that having a child is a private matter, the greater the chances of childbirth intentions being fulfilled.Footnote 22

Based on these results, we conclude that the specific features of the macro-social environment play a considerable role in the realization of fertility intentions. The model that includes inflation (uncertainty) and prevalent ideas related to individuality in childbearing proved the most promising; however, given the limitations of our analysis, caution is warranted.

As mentioned, we were aware that there were some doubts concerning the reliability of the estimates, due to the small number of countries (Level 2 cases), and we therefore carried out two kinds of robustness analysis. In order to determine whether any single country (especially a large one) could have a big influence on the coefficients, estimates were made omitting each country in turn. The results thus obtained show that none of the effects of the individual factors estimated for any 10 countries appears to differ from our estimate for all 11 countries (see Fig. 3 in the Appendix). The same analysis was carried out for the macro-level factors. In this case, too, the estimates obtained through the omission of each country in turn are clearly in line with the 11-country estimate presented in the paper (see Fig. 4 in the Appendix). This supports the robustness of the estimate for 11 countries: whichever country we leave out, the estimate never deviated significantly from the model estimated for the full sample (in all the models the control variables were retained).

Validity was also checked by estimating the realization probabilities for each macro factor for a specific social group (a female, aged 29–33, with a cohabiting partner, Parity 1, medium level of education, unemployed, making ends meet easily, medium subjective norm level) separately for each country.Footnote 23 The country-specific realization probabilities of the social group indicated (vertical axis) and the values of the macro variables (horizontal axis) and their linear associations are presented in individual graphs (see Fig. 5 in the Appendix). The plots show clearly along in which those cases where the estimated country probabilities fit smoothly along the estimated regression line and those cases where they are widely scattered. Our results are consistent with the findings of the multilevel models. The sign of the Pearson’s correlation coefficient was as expected in all cases. The relationship was not significant in all cases, despite the marked association, but this is basically due to the small number of cases (N = 11).

Discussion and Future Research

We are aware of the limitations of our study. Increasing the number of countries and the heterogeneity of the countries would certainly improve the models. We are also concerned about the high attrition rate in three countries, although our panel continuity analysis suggests that there are probably no strong biases due to the unusual attrition rates. As for the country-level variables used, new types of macro-level indicators may provide additional evidence. Despite the limitations, our analysis still enables us to provide some fresh insights into the study of fertility intention realization.

First, by taking advantage of the individual-level follow-up nature of the comparative GGS data, and by employing a rigorous approach to create the dependent variable, we have shown that there are considerable country differences in the realization of short-term intentions in Europe. The country rates for the overall probability of actually having a child, as intended, range from 16% (Bulgaria) to 40% (Germany, France).Footnote 24

Our results, which show a high degree of country heterogeneity in the realization of short-term intentions, are consistent with the results of earlier studies based on individual follow-up surveys. Studies comparing Western and Eastern European countries all show higher realization rates for Western European countries (Bradurashvili et al., 2011; Spéder & Kapitány, 2014). However, our estimation results based on individual data do not support estimates based on cross-sectional data (Harknett & Hartnett, 2014). The latter estimates arrive at higher realization rates than ours, and the ranking of countries in terms of realization differs from our estimates. As we have pointed out, this is mainly due to the fact that the country estimates based on cross-sectional data could not take account of cross-country heterogeneity in unplanned and sooner-than-intended childbearing (Brzozowska et al., 2021).

Second, based on our modelling, we conclude that specific features of the macro-social conditions play a considerable role in the realization of short-term fertility intentions. Exploring the country-level conditions using multilevel binary logistic regression with individual-level controls, we considered three major dimensions of likely influence. As far as economic conditions are concerned, the unemployment rate has been shown to be one of the most important macro-level factors determining fertility development (Goldstein et al., 2013), but we found that the realization of short-term fertility intentions is not influenced by the unemployment rate. Rather, it was the ‘swing’ in the youth unemployment rate that was significant, as well as the inflation rate: the less marked the swing in unemployment, and the lower the inflation rate, the greater the chances of fertility intentions being realized. The analysis of economic conditions highlights the fact that it is appropriate to start the study of realization with the well-established economic factors that influence fertility; but we have kept in mind that economic (and social) conditions already play a role in shaping intentions, and when looking for inhibiting or supporting conditions of realization, new types of economic factors should be considered in addition to the known ones.

We assumed that the type and level of socio-political commitment could contribute to the realization of intentions, since this indicates the availability of institutional resources for those in need. Two closely connected variables—social protection in general, as a percentage of GDP, and spending on children as a percentage of GDP—had a positive association with the likelihood of realization. This underscores the importance of the safety net in the realization of intentions. Nevertheless, in future research it would be useful to include more specific comparative indicators of government involvement: namely, comparative indicators to measure the stability of institutions or the introduction of new family policy measures.

We also considered the ideational condition of societies. Using three indicators, we demonstrated that ideational conditions play a role in the rate of realization. Beliefs about the private nature of the decision to have a child proved the most significant: the more support there was for the idea that ‘People should decide for themselves to have children’, the greater was the chance of respondents having the intended child. In other words, in societies with a weaker belief that having a child is also a public matter, people are less likely to ‘overstate’ their fertility intentions—or perhaps less likely to be overoptimistic about controlling the circumstances of having a child.

Our modelling does not help us arrive at a clear position on how combined macro-level effects that are individually significant would affect the realization of intentions. The applicability of our analytical strategy—the use of multilevel analysis—has already been bolstered by a number of additional analyses; and the inclusion of several macro variables in one model would, in our view, jeopardize our analytical approach. Therefore it remains the task of future research to clarify the link between macro factors that influence the realization of intentions; in the meantime, we propose to (at least provisionally) accept factors representing each domain that are independently significant.

The statistical analysis does not help us answer unequivocally the question of how macro-level factors influence the realization of fertility intentions; but considering the TPB framework, we can suggest some possible ways. In general, structural factors (labour market dynamics, inflation) are expected to directly hinder the realization of fertility intentions (as actual constraints): for example, inflation may increase the cost of having children or disrupt the expected access to housing. However, these factors can also lead to the revision of intentions, resulting in the initial intention not being realized. Less-generous welfare spending and family support provide less overall security in time of need and may also make individuals feel unsupported in their decision making. And so any unexpected change may break the link between intention and outcome. Finally, the attitudes and beliefs prevalent in society may lead to exaggerated intentions; and when these intentions come face to face with reality, they may be revised. Overall, these are assumptions: future research may develop approaches to examine the mechanism by which macro-level conditions influence the realization of intentions.